# Cross-lingual translation

Ultravox is a large multimodal speech language model that combines a pre-trained large language model and a speech encoder, capable of handling both speech and text inputs.

A multilingual reinforced language model fine-tuned based on the Qwen/Qwen2.5 - 32B - Instruct base model, proficient in multilingual processing and dealing with complex logical problems.

A multilingual sequence-to-sequence model based on Flan-T5-Large, supporting Vietnamese, English, and Chinese, suitable for tasks such as summarization, translation, and question answering.

WSPAlign is a weakly supervised large-scale span prediction-based word alignment pre-training model that supports word alignment tasks for multiple language pairs.